Conventional camera optics are derived from the eyes of vertebrates, wherein a single lens system captures light through a large aperture and focuses it onto a concave retina. Single aperture optics have the advantage of good spatial resolution and efficient light capture, but they are disadvantaged by their relatively large size and limited field of view. Vertebrates evolved to overcome the field of view limitation by gimbaling the eye in its socket, and by restricting the high resolution capability to the fovea centralis. Camera developers have attempted to extend the high resolution capability beyond the fovea by introducing additional optical elements to reduce the distortion that results from focusing an inherently concave image onto a flat film surface. The additional optical elements yield a lens that is expensive, heavy, and long. An example of a high resolution, wide angle lens is described by Momiyama in U.S. Pat. No. 4,437,735. The lens extends 20 inches beyond the image plane and uses 13 powered optical elements of different sizes, shapes, and materials.
Another disadvantage of the high resolution, single aperture lens is its need for focus adjustment to image objects at different distances. The problem is especially acute at close range and has prompted inventors to adopt various schemes to automate the focus adjustment process. One example of an auto focus system is described by Watanabe et al. in U.S. Pat. No. 7,184,090. It engages in a focusing operation while sending out an image capturing signal. It then settlers on the focus position that achieves the highest value in image contrast. As with all such auto focusing schemes, it cannot overcome the inherent design limitation of a single aperture lens: objects at different depths of field cannot be brought into focus simultaneously.
There is the need in many autonomous surveillance and robotic navigation applications for a distortion free, wide angle imaging system that remains in focus through all depths of field. Such a system could be modeled after the most popular eyes found in nature, the multiple aperture compound eyes of arthropods (i.e. insects and crustaceans). Compound eyes are formed from a convex array of micro-lenses (or lenslets) that collectively capture light through a very large field angle. The inherent advantage is that each sector of the field is separated into tiny zones that are imaged independently through lenslets positioned in the direction of the incoming image light. Since the aperture diameter and field angle of each lenslet are small, the corresponding optical aberrations are small. The composite image generated by the array is distortion free and remains in focus at all depths of field because each lenslet captures a very small section of the optical wavefront emanating from the object. The smaller the wavefront sections, the flatter they become until all objects appear to be at infinity. This is why arthropods have no need for a focusing mechanism.
Natural compound eyes can be divided into two general categories: apposition and superposition. In the apposition eye a simple lenslet focuses light directly onto a nearby receptive rod called a rhabdom. The two components constitute an ommatidium, of which there are thousands. Only a small cone of light along the axis of each ommatidium is detected. Light entering from outside the cone angle is absorbed in surrounding pigment cells. The spherical layout of the array enables adjoining lenslets to view adjacent fields. Though each lenslet image is inverted, in mosaic form the composite image appears erect because the lenslet viewing sectors are so small.
The architecture of the superposition eye varies slightly from that of the apposition eye. The superposition eye includes a meniscus shaped shell of long crystalline cones, a clear zone, and a convex rhabdom layer separated from the cones by a distance equal to half the radius of curvature of the outer meniscus surface. The cornea of each crystalline cone focuses incoming light within the cone and then collimates it in the latter part of the cone. The cone therefore acts as both a Keplerian telescope objective and an afocal eyepiece. Since the array of cones form a meniscus structure, the collimated light of a common field angle converge from adjoining cones to a single point on the confocal contour of the rhabdom layer. Thus the light from all of the cones separate according to field angle and then superpose on the rhabdom surface to produce a single, upright image. Since light from a large number of cones contribute to each field point in the image, the effective sensitivity of a superposition eye is increased relative to an apposition eye. This is why apposition eyes are found primarily on diurnal arthropods, such as butterflies, and superposition eyes are found primarily on nocturnal arthropods, such as moths.
Despite the inherent advantages of the superposition architecture, artificial compound eyes are more commonly derived from the much simpler apposition format. Duparré et al. describe a flat lenslet array artificial compound eye the size and shape of a credit card (see Duparré et al., Applied Optics, August 2004, pp 4303-4310, vol. 43, No. 22). The flat design attribute is beneficial in that it enables the use of flat lenslet arrays, which are readily manufactured in a variety of ways (see for example Fadel et al., U.S. Pat. No. 6,967,779). The flat design also matches well to flat mosaic detector arrays, which are easy to manufacture and readily available. However, the flat design attribute limits the field of view to just 21 degrees.
Another variation of a flat lenslet array system is described by Gurevich et al. in U.S. Pat. No. 7,187,502. This system uses a second flat lenslet array of a different pitch to increase the magnification of the image. The system was invented for imaging “remotely located objects, i.e., objects located behind the focal distance of the assembly”. Though the flat lenslet arrays described in these inventions are readily available, a curved lenslet array would enable a larger field of view and allow the optics to be made conformal to its mounting structure.
Lee and Szema describe an artificial apposition array compound eye that closely mimics the design found in nature (see Lee and Szema, Science, November 2005, pp 1148-1150, vol. 310, No. 5751). The lenslet array is convex in shape, and the light from each lenslet is focused onto a convex surface. Unfortunately, the design requires a convex shaped detector array of extremely small size to capture the image.
Another apposition compound eye concept is described by Sweatt and Gill in U.S. Pat. No. 7,286,295. In this concept the lenslets have power on two surfaces and are preferred to be aspheric to correct for optical aberrations. The lenslets are made from polymethyl methacrylate and are optimized for a single wavelength. The lenslets are separated laterally along the array by a spacer baffle that prevents cross-talk between cells. The lenslets focus a series of inverted sub-images onto a dome shaped, coherent fiber optic bundle that is supposed to transport the sub-images onto a flat detector array. However, the fiber optic bundle is not tapered, and so the fiber tips are beveled along the peripheral regions of the dome. The bevels prevent light capture along the axes of the peripheral lenslets and encourage stray light capture from oblique angles. Gaps and overlaps in the sub-images are controlled by the spacing of the lenslets and the curvature of the array. Since the sub-images are each inverted, the composite image must be constructed digitally by post-processing.
An artificial superposition array compound eye is described in U.S. Pat. No. 7,376,314. In this concept two lenslet arrays are hot press molded into a convex, meniscus form. The lenslets are paired to operate as afocal Keplerian telescopes that focus, collimate, and bend the incoming light. The meniscus form enables the collimated light from adjacent lenslets to be directed toward a common point on the convex surface of a fiber optic imaging taper. In this manner all of the lenslets work together to form a single, upright, high intensity image on top of the taper. The taper transfers the upright image to a flat detector array; no digital post-processing is required. The fiber tips of the taper are each cut perpendicular to the fiber axes, so only image light from the correct angles are captured by the fibers. A honeycomb louver baffle is positioned between the lenslets and the taper dome to block ghost images. The diameter of the honeycomb cells sets the effective pupil size of the optics. A typical cell diameter encompasses 100 lenslets out of the 30,000 lenslets in each array, thereby increasing its sensitivity by a factor of 100 over an equivalent apposition eye.
The main shortcoming in the artificial superposition eye is its spatial resolution, which is limited by optical aberrations and the diameter of the collimated beams that overlap to form the image. The beam diameters can be reduced by decreasing the diameter of the lenslets in the array. However, if the lenslets are to be manufactured by hot press molding, then there are practical limits imposed on the minimum lenslet diameter and on the alignment accuracy between the two arrays. Hot press molding tends to shift the lenslets on the outer convex surface toward the array center, and it tends to shift the lenslets on the inner concave surface away from the array center. This causes blurring of the image. Though the afocal optical aberrations can be corrected by introducing more lenslets arrays and different glass types, it is difficult to accomplish this in a superposition architecture, especially with regard to lenslet alignment if the lenslet diameters are very small.
The main shortcoming in the artificial apposition eye is its sensitivity, which is limited by the diameter of the individual lenslets. If the lenslet diameters are increased, then each lenslet must cover multiple field angles, thereby creating sub-images. Since the sub-images are inverted, a coherent image can only be constructed by post-processing; reconstruction cannot be accomplished where the sub-images overlap on the detector array. It is therefore better to design the optics and baffles to leave gaps between the sub-images while generating redundant image features along the edges of adjacent sub-images. It is then possible to crop the redundant regions when the sub-images are spliced together digitally.
The ideal artificial compound eye is a hybrid that combines the sensitivity of the superposition eye with the resolution of the apposition eye. It generates a single, upright image without the need for post-processing. Since hybrid compound eyes are not found in nature, the design architecture requires invention from first principles.
U.S. PATENT DOCUMENTS4,357,088Nov. 2, 1982Pomerantzeff4,437,735Mar. 20, 1984Momiyama6,449,103Sep. 10, 2002Charles6,571,043May 27, 2003Lowry et al.6,898,015May 24, 2005Yoshikawa et al.6,933,167Aug. 23, 2005Yamamoto6,942,959Sep. 13, 2005Dubin et al.6,967,779Nov. 22, 2005Fadel et al.7,106,529Sep. 12, 2006Kerr et al.7,119,962Oct. 10, 2006Gurevich et al.7,184,090Feb. 27, 2007Watanabe et al.7,187,502Mar. 6, 2007Gurevich et al.7,217,573May 15, 2007Oshida et al.7,286,295Oct. 23, 2007Sweatt et al.7,376,314May 20, 2008Reininger